Add functions for fast bulk conversion of qfloat16

Both ARM and x86 can convert fp16 much faster in bulk than one at a
time. This also enables hardware accelerated conversion on x86, when
F16C isn't unconditionally available at compile time.

This code is implemented in C to ensure that there's no leakage of
inline symbols from the .obj file that was compiled by Visual Studio
with AVX support. Unfortunately, simd.prf uses $(CXX) instead of $(CC)
for all its sources, which means the file gets interpreted as C++ by
g++, clang++ and icpc. Those compilers at least don't leak any symbols.

Done-with: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I9d26d99e83392861fb09564e0e8e8d76cd8483b3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

2 files changed, 44 insertions, 0 deletions

diff --git a/tests/auto/corelib/global/qfloat16/tst_qfloat16.cpp b/tests/auto/corelib/global/qfloat16/tst_qfloat16.cpp
index 9bd87e3f21..d2902f4944 100644
--- a/tests/auto/corelib/global/qfloat16/tst_qfloat16.cpp
+++ b/tests/auto/corelib/global/qfloat16/tst_qfloat16.cpp
@@ -46,6 +46,8 @@ private slots:
     void promotionTests();
     void arithOps_data();
     void arithOps();
+    void floatToFloat16();
+    void floatFromFloat16();
 };
 
 void tst_qfloat16::fuzzyCompare_data()
@@ -305,5 +307,41 @@ void tst_qfloat16::arithOps()
     QVERIFY(qFuzzyCompare(r4,1.f/val2));
 }
 
+void tst_qfloat16::floatToFloat16()
+{
+    float in[63];
+    qfloat16 out[63];
+    qfloat16 expected[63];
+
+    for (int i = 0; i < 63; ++i)
+        in[i] = i * (1/13.f);
+
+    for (int i = 0; i < 63; ++i)
+        expected[i] = qfloat16(in[i]);
+
+    qFloatToFloat16(out, in, 63);
+
+    for (int i = 0; i < 63; ++i)
+        QVERIFY(qFuzzyCompare(out[i], expected[i]));
+}
+
+void tst_qfloat16::floatFromFloat16()
+{
+    qfloat16 in[35];
+    float out[35];
+    float expected[35];
+
+    for (int i = 0; i < 35; ++i)
+        in[i] = qfloat16(i * (17.f / 3));
+
+    for (int i = 0; i < 35; ++i)
+        expected[i] = float(in[i]);
+
+    qFloatFromFloat16(out, in, 35);
+
+    for (int i = 0; i < 35; ++i)
+        QCOMPARE(out[i], expected[i]);
+}
+
 QTEST_APPLESS_MAIN(tst_qfloat16)
 #include "tst_qfloat16.moc"
diff --git a/tests/auto/corelib/global/qglobal/qglobal.c b/tests/auto/corelib/global/qglobal/qglobal.c
index 6a6f53dfe0..bf2943a63c 100644
--- a/tests/auto/corelib/global/qglobal/qglobal.c
+++ b/tests/auto/corelib/global/qglobal/qglobal.c
@@ -71,6 +71,12 @@ void tst_GlobalTypes()
 
     qreal qr;
     Q_UNUSED(qr);
+
+    qssize_t qs;
+    qptrdiff qp;
+    qintptr qip;
+    quintptr qup;
+    Q_UNUSED(qs); Q_UNUSED(qp); Q_UNUSED(qip); Q_UNUSED(qup);
 }
 
 /* Qt version */